Better Than The Models ?

[bluearmy]

i was 'critical' (i thought honest) re the last comparison chart. today's looks like a pretty fair result. not perfect by any means but reasonable. be interesting to see what roger evaluates it at.

[AtlanticFlamethrower]

12z ECM doesn't look a match-up at all. The big low stays West of Iceland and then goes NW. MB needs it East of Iceland, preferably going SE.

[TEITS]

12z ECM doesn't look a match-up at all. The big low stays West of Iceland and then goes NW. MB needs it East of Iceland, preferably going SE.

I don't mean to be rude but I don't know why some are still looking at the model output to see if MB forecast is going to be correct because like I said earlier it just ain't going to happen.

[Roger J Smith]

One experimental forecast is making a weak rally today but it starts with R and not M.

Anyway, the day 4 results are in. I doubt that these summaries will change much now because the two main models (GFS, ECM) are locked into solutions that are correlated above 0.9 both with each other and previous day's runs. It would be highly unusual for this kind of consensus to disappear into a totally different solution within 96 hours. Such minor changes as have occurred are actually boosting the RJS correlation by a few points (I think this is because the solutions are more moderate than yesterday on gradient and the RJS map was a weaker zonal breakdown sort of pattern that might have correlated quite well with tomorrow's map).

From today on, I will not comment on the maps themselves in any detail because there are only slight changes and other posters have covered these aspects earlier. Suffice it to say that the GFS has weakened slightly the overall gradient between low pressure east of Iceland and high pressure ridged east-west near 47N, while ECM holds off the low to the west of Iceland and continues a rather moderate gradient into the British Isles, North Sea and southern Scandinavia. Hints of deeper low pressure in north-central Scandinavia on earlier runs are fading out.

The GFS day 4 is a virtual copy of the previous day with correlation 0.97. It has dropped in correlation with the MB forecast map to 0.12, and shown a slight gain vs RJS to 0.44.

The ECM day 4 correlates at 0.94 with its previous day run and at 0.92 with the GFS. It has now dropped back to a slight negative correlation with MB (due to the filling of pressures over Sweden, I believe) at -0.15 today, while inching up to 0.43 vs RJS.

The separate low west of Ireland that is a feature on the RJS map is represented more by a trough within the fast WSW flow on both of the main models.

Look for brief summaries of days 3, 2 and 1 here but these are probably going to be a series of slight adjustments to the correlation numbers. Starting Tuesday I also have the larger validation system to check as we compare MB and RJS predictions (the latter in post 190) against real time outcomes.

It's beginning to look as though MB will have to consider a combination of broad success and failure paradigms in his forecast. The broad success would be to identify a period of faster flow and larger pressure gradients than we've seen most of the winter so far. The broad failure would be in terms of the flow being at a fairly large angle to what was envisaged (NW as opposed to WSW), and some will also say the intensity although I think part of that may be subjective, whereas what we're comparing here are objective statistics from the map itself.

In my own case, I think the map is probably what would verify at fairly high correlation if the upper level winds at map time were about two-thirds of actual strength, with fairly slight adjustments towards the Baltic, so in that sense, the experimental result is pointing at the strength of blocking as the main problem in that forecast. There may be some room for improvement in the correlation outcome seeing that the stats have been inching up for two days now (through what I would define as moderately high range for this sort of effort).

I would encourage anyone out there with ideas of how to create maps for over 30 days out to start thinking about this date: 21 March at 12z, as I will invite any interested parties to submit a map for that date in a thread to be started after this validation study ends, to be submitted no later than 19 Feb (30d in advance).

[MurcieBoy]

Evening folks, thanks very much RJS for the update evaluation. Here are the 30 Jan 12z charts for GFS and ECM:

1 Feb

2 Feb

3 Feb

4 Feb

5 Feb

Actuals:

28 Jan

30 Jan

Edited January 30, 2011 by MurcieBoy

[AtlanticFlamethrower]

The deep low on the 18z has shifted very slightly East, directly north of Scotland. It now has a double core. The lower core swings furthest South and East. This double core low solution might help MB get it further South and East.

Looking at 28 / 30 Jan Actual versus predicted, that's not bad if we are being uncharitable and assuming MB is making a complete guess, for his own amusement at our expense.

[Timmytour]

I don't mean to be rude but I don't know why some are still looking at the model output to see if MB forecast is going to be correct because like I said earlier it just ain't going to happen.

Presumambly you are saying that goes for both the MB forecast AND the model output?

[Coast]

Here was yesterdays MB forecast milestone on the applicable nearest fax chart:

There is no chart or comment for today so the next one to compare will be tomorrows:

[AtlanticFlamethrower]

Looks like the winds are going to come from the wrong direction on 3rd of February - SWly - but NAO +ve is right and these other charts would be reasonably helpful to someone starting to think about making a long range forecast.

Not long before we know how the final correlations of RJS and MB stack up.

[MurcieBoy]

Morning folks, re the winds, just to clarify:

[Coast]

Gradient wind

A hypothetical wind based upon the assumption that the sum of the horizontal components of the Coriolis force and the atmospheric pressure gradient force per unit mass is equivalent to a wind acceleration which is normal to the direction of the wind itself (centripetal acceleration), with the implication that there are no viscous forces acting. The direction of the gradient wind is the same as that of the geostrophic wind. The gradient wind speed is less than the geostrophic speed when the air moves in a cyclonically curved path and greater when the air moves in an anticyclonically curved path. The gradient wind is a good approximation of the actual wind and is often superior to the geostrophic wind, particularly when the flow is strongly curved in the cyclonic sense.

http://www.answers.com/topic/gradient-wind#ixzz1CbNx73FR

http://ww2010.atmos.uiuc.edu/(Gh)/guides/mtr/fw/grad.rxml

http://www.cartage.org.lb/en/themes/sciences/earthscience/hydrology/meteorology/BalancedAtmospheric/GradientWind/GradientWind.htm

Deriving Wind Speed

Performing exact wind speed calculations means determining acceleration, for which you must have the density of the air, and then converting it to speed. This involves some rather complicated equations and generally the use of an assumed average air density that is not always correct in individual cases. Thankfully, using contour intervals, estimates can be derived. "How to Read and Interpret All the Basic Weather Charts" states: "A pressure difference of 4 mb every 60 miles would give winds of about 50 mph." Considering that change in pressure gradient exhibits a positive and linear relationship with speed, a 4-mb difference every 30 miles would bring approximately 120-mph winds

http://www.ehow.com/about_5100646_can-isobars-tell-wind-speed.html#ixzz1CbPQxYgS

[Alan Robinson]

http://www.answers.c...d#ixzz1CbNx73FR

http://ww2010.atmos....tr/fw/grad.rxml

http://www.cartage.o...radientWind.htm

http://www.ehow.com/...l#ixzz1CbPQxYgS

I am always astounded when people speak of the Coriolis Force as though it actually exists. It doesn't.

The Coriolis Effect certainly exists, and is a consequence of coordinate reference systems moving relative to each other. Imagine a missile fired due north from the equator, remembering that north / south simply means the alignment of a great circle passing through the poles. Now relative to solar system coordinates - however they are defined - the velocity of a missile fired from the equator towards the north pole has also a considerable west-to east component because of the Earth's rotation - which is greatest at the equator and zero at the poles - and this equatorial west to east component becomes increasingly evident as the missile approaches higher latitudes. Put another way, people on the equator are moving through the solar system faster than those in Greenland. For an observer on Earth, it appears that the missile's path assumes an eastward curvature. This is the Coriolis effect. Engineers dealing with the Coriolis effect must correct for the apparent drift if they want to hit the pole, and this correcting force corresponds to what is called the Coriolis force.

The Coriolis force in fact is a theoretical force used to build mathematical models for correcting courses. There is no real-life Coriolis force, and it astounds me to read explanations of physical phenomena that refer to the Coriolis force. I even heard a load of rubbish being spoken about large tank ships having to correct for the Coriolis force as they cross the oceans. What a load of tosh; if their course is unsteady it is because of directional instability (wrong trim), propeller forces, rudder slightly misaligned, or unequal marine growth on the hull. Or have I misunderstood?

Edited January 31, 2011 by Alan Robinson

[Roger J Smith]

Day 3 correlation report

_______________________________

There is very little change in either the GFS or the ECM over the past 24h, so yesterday's comments still apply. If anything, the ECM has shifted slightly towards the more consistent (in this case) GFS, with the position of the Iceland low and a hint of a secondary WNW of Ireland. This is a more intense feature than the entirely separate low on the RJS model (for both GFS and ECM). The trend of the past several days has been to maintain a weak low pressure centre near the MB low in north-central Sweden but rather than being a powerful feature this is just a remnant of earlier low pressure that moved into the region and filled.

To give an idea of which major model is more consistent, day 3 correlates at only .05 with day 10 for the ECM but at a more robust .59 for the GFS. In other words, the GFS has been much more consistent through the period leading up to real time. We'll see soon if this indicates a firm prediction of a real solution.

Just to give some idea of how these correlation stats have been trending, I am going to list them all from day 10 to the present (day 3) and then update this chart each day to day zero.

statistic ...... DAY 10 .. DAY 09 ...DAY 08 .. DAY 07 ...DAY 06 .. DAY 05 .. DAY 04 .. DAY 03 ..

GFS

day to day ..... ----- ...... 0.87 ...... 0.77 ...... 0.84 ...... 0.90 ...... 0.92 ...... 0.97 ...... 0.96

GFS

vs MB .......... 0.15 ....... 0.05 ...... 0.06 ......-0.22 ...... 0.08 ...... 0.19 ...... 0.12 ......-0.07

GFS

vs RJS ......... 0.53 ....... 0.36 ...... 0.60 ...... 0.56 ...... 0.52 ...... 0.40 ...... 0.44 ...... 0.39

ECM

day to day ..... ----- ......-0.19 ...... 0.59 ...... 0.64 ...... 0.85 ...... 0.88 ...... 0.92 ...... 0.98

ECM

vs MB ..........-0.79 .......-0.15 ......-0.20 ...... 0.48 ...... 0.32 ...... 0.12 ......-0.15 ......-0.09

ECM

vs RJS ......... 0.07 ....... 0.71 ...... 0.31 ...... 0.17 ...... 0.36 ...... 0.37 ...... 0.43 ...... 0.38

ECM

vs GFS .........-0.37 ....... 0.65 ...... 0.55 ...... 0.81 ...... 0.94 ...... 0.96 ...... 0.92 ...... 0.99

-------------------------------------------------------------------------------------------

Comments or questions welcome.

Edited January 31, 2011 by Roger J Smith

[MurcieBoy]

As always, thanks RJS. Here are the 31 Jan 12z runs:

1 Feb

2 Feb

3 Feb

4 Feb

5 Feb

Actuals:

28 Jan

30 Jan

[AtlanticFlamethrower]

This is GFS chart for Feb 4, 6am.

If that could come 12hr earlier and a little further south MB's Feb 3rd forecast might yet look accurate in detail, (as opposed to accurate in "NAO +ve pattern broadly correct", which is easier to do).

An even better chart for MB was on the 12z GEM

12am, Feb 4th.

Game back on?

[Timmead]

I am always astounded when people speak of the Coriolis Force as though it actually exists. It doesn't.

The Coriolis Effect certainly exists, and is a consequence of coordinate reference systems moving relative to each other. Imagine a missile fired due north from the equator, remembering that north / south simply means the alignment of a great circle passing through the poles. Now relative to solar system coordinates - however they are defined - the velocity of a missile fired from the equator towards the north pole has also a considerable west-to east component because of the Earth's rotation - which is greatest at the equator and zero at the poles - and this equatorial west to east component becomes increasingly evident as the missile approaches higher latitudes. Put another way, people on the equator are moving through the solar system faster than those in Greenland. For an observer on Earth, it appears that the missile's path assumes an eastward curvature. This is the Coriolis effect. Engineers dealing with the Coriolis effect must correct for the apparent drift if they want to hit the pole, and this correcting force corresponds to what is called the Coriolis force.

The Coriolis force in fact is a theoretical force used to build mathematical models for correcting courses. There is no real-life Coriolis force, and it astounds me to read explanations of physical phenomena that refer to the Coriolis force. I even heard a load of rubbish being spoken about large tank ships having to correct for the Coriolis force as they cross the oceans. What a load of tosh; if their course is unsteady it is because of directional instability (wrong trim), propeller forces, rudder slightly misaligned, or unequal marine growth on the hull. Or have I misunderstood?

A bit off topic, but it is a force as far as most physicists are concerned (albeit a "fictitious" one in that it does, as you say, originate from relative rotation of different co-ordinate reference frames), as it must be considered in dynamics calculations in a rotating frame. For example, a stone dropped perfectly vertically from a helicopter at the equator in totally calm conditions will experience the Coriolis force and thus drift slightly eastwards. If you ignored it, experiments wouldn't match your theory. Therefore it is a "real" force

Edited January 31, 2011 by Timmead

[MurcieBoy]

As the GFS and ECM models are converging and GFS comes out earlier, will only compare Storm Forecast with GFS. This is today's 06z GFS comparisons:

1 Feb

2 Feb

3 Feb

4 Feb

5 Feb

Actuals:

28 Jan

30 Jan

[Coast]

12Z Met fax chart analysis with MB's forecast for 14.00hrs superimposed over the top:

I can't find any areas of pressure that correspond, isobars over the UK and into the North Sea are at approximately 70 to 90 Dg to each other (actual to forecast) and no fronts correspond. To my untrained eye, I'd say today was not a good one.

Positive news for the outlook is there is a substantial low showing up for the forecast period which could bring high winds and storms to Northern Scotland. Negative side of that is it appears to be coming from Canada via the Atlantic.

http://www.metoffice.gov.uk/weather/marine/highseas_forecast.html#All~All

[Roger J Smith]

UPDATE on validation study ... at day 2 (48h) from map time

_____________________________________________________________

The global models have changed very little in the past 24 hours as shown by their high correlations (see table below). The ECM does for the first time identify a separate low northwest of Ireland at about 57N 18W compared to 55N 15W on the RJS model. However this low is more intense and also a secondary of a deeper low near Iceland.

In terms of the larger-scale validation suggested in post 190, today would be day one of the five-day time window for verifying storm tracks, and maximum winds. Storm track will be difficult to judge since for MB, the low to be tracked is going to verify on day 3 but would not be moving much from today to then, whereas his series of maps shows a more mobile flow and would suggest an origin northeast of Greenland which is not going to be the case. In my own case, the low which will probably verify well on day 3 is a new development and if I tracked the wave energy there would be a similarly large error today. I will verify this by the numbers in both cases and score the day 1 positions at zero (larger than 700 km error). However this part of the validation is going to be eclipsed, I think, by considerations of overall intensity and flow pattern, jet stream direction etc. As the max wind validations were for any day in the period, I think we can assume today won't be that day. These may actually end up looking fairly good for MB's forecast (after I cut him that break on the 80 knot max surface gust and 100 knot max 500 mb). We shall see how that works out on the night of 3-4 Feb when at the moment I would expect these criteria to come fairly close to verifying (albeit from SW rather than NW).

On direction of the upper level flow, where I was proposing to score daily, the Shetlands direction at 12z was SW or about 230 deg. This will provide a small score for both of us with our guesses of NW and SE. I will get a table going on these validations by the third when there are some scores to report.

As to the ongoing correlation study, here is yesterday's chart updated for the 48h GFS and ECM output.

statistic ...... DAY 10 .. DAY 09 . DAY 08 .. DAY 07 . DAY 06 .. DAY 05 . DAY 04 .. DAY 03 . DAY 02

GFS

day to day ..... ----- ...... 0.87 ...... 0.77 ...... 0.84 ...... 0.90 ...... 0.92 ...... 0.97 ...... 0.96 ...... 0.99

GFS

vs MB .......... 0.15 ....... 0.05 ...... 0.06 ......-0.22 ...... 0.08 ...... 0.19 ...... 0.12 ......-0.07 ......-0.12

GFS

vs RJS ......... 0.53 ....... 0.36 ...... 0.60 ...... 0.56 ...... 0.52 ...... 0.40 ...... 0.44 ...... 0.39 ...... 0.44

ECM

day to day ..... ----- ......-0.19 ...... 0.59 ...... 0.64 ...... 0.85 ...... 0.88 ...... 0.92 ...... 0.98 ...... 0.97

ECM

vs MB ..........-0.79 .......-0.15 ......-0.20 ...... 0.48 ...... 0.32 ...... 0.12 ......-0.15 ......-0.09 ......-0.00

ECM

vs RJS ......... 0.07 ....... 0.71 ...... 0.31 ...... 0.17 ...... 0.36 ...... 0.37 ...... 0.43 ...... 0.38 ...... 0.38

ECM

vs GFS .........-0.37 ....... 0.65 ...... 0.55 ...... 0.81 ...... 0.94 ...... 0.96 ...... 0.92 ...... 0.99 ...... 0.97

---------------------------------------------

Although we need to complete the validation study, I think the conclusions are already fairly obvious and were stated in yesterday's post. Will be back same time tomorrow to do it all over again.

Edited February 1, 2011 by Roger J Smith

[shotski]

OK so its 1st of Feb and Murcieboys prediction is 100% verifying, 130 mile an hour winds and major destruction is on the cards

The only problem is its on the other side of the world in Australia

But joking apart if MB had come on here in November and predicted a destructive storm with 130 mile an hour winds would hit at this exact

time somewhere in the world that would be some prediction to come off.

Maybe he is on to something but just needs to tweet the data

The Nort Sea Storm (Australia)

[Paul Sherman]

Dont think Nick would stoop to the levels of Piers Corbyn Shotski!

Cant wait until Piers and His cronies accept the plaudits for the Heavy Snow We USA Are getting, But as they say that is for an entirely different thread.

[MurcieBoy]

OK so its 1st of Feb and Murcieboys prediction is 100% verifying, 130 mile an hour winds and major destruction is on the cards

The only problem is its on the other side of the world in Australia

But joking apart if MB had come on here in November and predicted a destructive storm with 130 mile an hour winds would hit at this exact

time somewhere in the world that would be some prediction to come off.

Maybe he is on to something but just needs to tweet the data

The Nort Sea Storm (Australia)

.... great post! Made me chuckle.

.... perhaps that was my "schoolboy error" - wrong hemisphere!!!!

EDIT: gosh, nothing funny about this

Edited February 2, 2011 by MurcieBoy

[Coast]

[MurcieBoy]

Morning, here are the Storm Forecast v GFS comparisons and actuals as at 6z today:

2 Feb

3 Feb

4 Feb

5 Feb

Actuals:

28 Jan

30 Jan

1 Feb

[Coast]

Here's my overlay of the two applicable charts for 02/02/11:

As yesterday, isobar spacing and relative direction are out in respect of each other and the fronts forecast, don't correspond to anything on the 12z analysis except for very loosely over Germany into the former USSR countries.

Today's interesting point is the location of the two low pressure areas around Iceland and Norway. Both the forecast ones are only a few hundred odd miles out of position to the actual, but the easterly one is 30mb lower at its centre than actual and the westerly one is 13mb higher.